Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Jian-Wen Bao (NOAA/ESRL/PSD) Sara A. Michelson (NOAA/ESRL/PSD)
S. G. Gopalakrishnan (NOAA/AOML/HRD) Frank Marks (NOAA/AOML/HRD)
Jun Zhang (NOAA/AOML/HRD)
Vijay Tallapragada (NOAA/NCEP/EMC)
Comparison and Evaluation of Two ABL Mixing Schemes in HWRF
Presented at The 66th Interdepartmental Hurricane Conference
Charleston, SC, 5-8 March 2012
Outline
1. Function of ABL parameterization schemes in NWP models 2. Summary of two ABL schemes in the HWRF model
3. Comparison in HWRF idealized intensification experiments 4. Preliminary conclusions
What does the ABL parameterization scheme do?
• Sub-grid turbulence transports temperature, moisture and momentum (+ tracers).
• Attempts to integrate effects of sub-grid scale turbulent motion on prognostic variables at grid resolution.
Operational measure for success: correct model output on the grid-resolved scales everywhere in the model domain. Scientific measure for success? Challenge: The simulated turbulent mixing is quite dependent on the definition of the ABL depth and structure!
Two ABL Schemes in the HWRF Model • The NMM ABL scheme (MYJ): K-theory, 1.5-oder turbulent kinetic energy (TKE) prognostic equation with non-singular realization
• The GFS ABL scheme: non-local closure with modifies K- profile of Troen and Mahrt (1986) where is the non-local flux representing the influence of the large-eddy mixing in the ABL with convection.
TKElKPKzCKwc
zuKwu
MMCrM
CM
α==∂∂
−=∂∂
−=
−1
'',''
21 )/1()/(
'''',''
hzhzhwKPK
wczCKwc
zuKwu
sCrM
nlCM
−==
+∂∂
−=∂∂
−=
− κ
''wcnl
Property Summary of the Two Schemes
1. The GFS scheme assumes that there is a well defined layer h in which the vertical distribution of diffusivities follows a special cubic polynomial function x(1-x)2, where x = z/h.
2. The MYJ scheme naturally links the ABL depth to the sub-grid TKE distribution, though it does require a specification of the mixing length.
3. The magnitude of the diffusivities from the GFS scheme is determined by the ABL depth, while in the MYJ scheme it is determined by the mixing length.
Table of Experiments
model grid spacing: dx = dy = ~9 km, ~ 3 km, kx = 43 (NMM sigma-p levels); operational microphysics and radiation schemes
Experiment Name ABL Scheme Surface Layer
Scheme
Convective Parameterization
Scheme
Profile of Vertical Diffusivity
MYJ/GFS MYJ GFS SAS on both grids Original
MYJ/GFS/NOCPS MYJ GFS None on both grids Original
MYJ/GFS/0.9alph MYJ GFS SAS on both grids Original, Increase ALPH to 0.9
GFS/GFS GFS GFS SAS on both grids Original
GFS/GFS/NOCPS GFS GFS None on both grids Original
GFS/GFS/0.5vertdiff8pfac GFS GFS SAS on both grids 0.5 Vertical Diffusion, Increase PFAC to 8
GFS/GFS/0.25vertdiff GFS GFS SAS on both grids 0.25 Vertical Diffusion
GFS/GFS/8PFAC GFS GFS SAS on both grids Original, Increase PFAC to 8
Conventional Intensity Comparisons
low bias with GFS + no convection
Comparisons of vertical eddy diffusivities with observations at 500 m AMSL
Left 2 rows: Azimuthally averaged tangential acceleration (color shaded, ms-1h-1), radial wind speed (red contours, contour interval 5 ms-1) and Km (black contours, contour interval 50 m2s-1). Right: Km profiles
MYJ GFS
GFS GFS
MYJ GFS noCPS
Eddy diffusivity profile and magnitude control the depth of BL inflow and tangential acceleration averaged over 96-108 h: MYJ vs GFS
MYJ Km Profile
GFS GFS nocps GFS Km profile
Impact of the magnitude of diffusivity on tangential acceleration and R-Z wind structure averaged over 96-108 h: GFS inter-comparison
GFS 0.25 v-diff
GFS GFS
Azimuthally averaged tangential acceleration (color shaded, ms-1h-1), radial wind speed (red contours, contour interval 5 ms-1) and Km (black contours, contour interval 50 m2s-1)
Azimuthally averaged tangential wind speed (color shaded, contour interval 5 ms-1 ) and radial wind speed (purple contours, contour interval 3 ms-1)
Km Profile
Impact of Zkmax on the depth of BL inflow and tangential acceleration averaged over 96-108 h: GFS inter-comparison
GFS 0.25v diff Km
GFS 8pfac Km
Left: Azimuthally averaged tangential acceleration (color shaded, ms-1h-1), radial wind speed (red contours, contour interval 5 ms-1) and Km (black contours, contour interval 50 m2s-1). Right: Km profiles
MYJ GFS
MYJ GFS SFCLAY 0.9 alph
Impact of the magnitude of diffusivity on the tangential acceleration and the R-Z wind structure averaged over 96-108 h: MYJ inter-comparison
Azimuthally averaged tangential acceleration (color shaded, ms-1h-1), radial wind speed (red contours, contour interval 5 ms-1) and Km (black contours, contour interval 50 m2s-1)
Azimuthally averaged tangential wind speed (color shaded, contour interval 5 ms-1 ) and radial wind speed (purple contours, contour interval 3 ms-1)
Km Profile
MYJ GFS GFS GFS GFS GFS 0.5v diff 8pfac
GFS with MYJ-like diffusivity profile vs GFS profile averaged over 96-108 h
Top: Azimuthally averaged tangential acceleration (color shaded, ms-1h-1), radial wind speed (red contours, contour interval 5 ms-1) and Km (black contours, contour interval 50 m2s-1). Bottom: Km profiles
Preliminary Conclusions 1. The intensification rate is influenced by both the profile and magnitude
of the vertical eddy diffusivity.
2. Comparisons with limited observations appear to be in favor of the MYJ scheme in terms of the magnitude of diffusivities, but we do not have conclusive results for the vertical profile.
3. The simulated intensification using the GFS scheme is more sensitive to the use of the SAS convection scheme than the MYJ scheme.
4. The R-Z structure of the simulated TC in terms of the enclosed area of 55 m/s becomes broader as the magnitude of the eddy diffusivities increase in both schemes, but the MYJ scheme produces a shallower BL inflow than the GFS scheme.
5. When the GFS scheme is used, the tangential acceleration above the ABL inflow increases as the magnitude of eddy diffusivities decrease, while it is opposite when the MYJ scheme is used.
6. It is possible to change the profile and magnitude of the GFS diffusivities to mimic those of the MYJ scheme.